Functional ADL


func-adl is an implementation of a functional analysis description language that is declarative. The project is exploring several different aspects of this:

  • The separation between a query language and a complete analysis
  • Can a query be reduced to a single abstract syntax tree or DAG computational graph
  • Can multiple user-interface languages use the same backend?
  • Given that physics is usually to be evaluated in a per event basis, can the user analysis be written in such a way that the user thinks only about single event quantities even though the manipulation is done in a columnar way?
  • Can a common query language be used to run on different data formats (binary experimental format, flat ROOT TTree’s, and awkward array columnar analysis)

Status

This project is currently in production for ServiceX. It works with the xAOD and uproot back ends.

GitHub Repositories

Repositories connected with this work:

  • func_adl - The base package for the func_adl SQL-like data query language.
  • func_adl_xAOD - Backend that works with ATLAS xAOD files to extract columnar data.
  • func_adl_uproot - Backend that works with ROOT TFile’s that contain TTree’s.
  • qastle - Interchange language that converts func_adl to a text language suitable for transmission over the web. While list-like and plain-text, it is meant as an interchange format, not an end-user language.

This work is based on the original work done with LINQ in the C# programming language (prior to the start or IRIS-HEP):

Team